github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-06 08:02:14 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #4904] Need Support: Local Model Parameters Override Like Llama.cpp #3098

New Issue
Open
opened 2026-04-12 13:32:25 -05:00 by GiteaMirror · 4 comments
GiteaMirror commented 2026-04-12 13:32:25 -05:00
Owner
Copy Link

Originally created by @DirtyKnightForVi on GitHub (Jun 7, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/4904

In llama.cpp, when running a model, I can update the model parameters using --override-kv.

How can this be achieved in ollama?
Should I modify a certain file?
Or PARAMETER to the Modlefile?
Or is there some other similar command?

Here is a situation that i have to override some parameters.
It works at llama.cpp, but i am not sure in ollama.

Metadata KV overrides (pass them using --override-kv, can be specified multiple times):

deepseek2.attention.q_lora_rank=int:1536
deepseek2.attention.kv_lora_rank=int:512
deepseek2.expert_shared_count=int:2
deepseek2.expert_feed_forward_length=int:1536
deepseek2.expert_weights_scale=float:16
deepseek2.leading_dense_block_count=int:1
deepseek2.rope.scaling.yarn_log_multiplier=float:0.0707
Originally created by @DirtyKnightForVi on GitHub (Jun 7, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/4904 In llama.cpp, when running a model, I can update the model parameters using `--override-kv`. How can this be achieved in ollama? Should I modify a certain file? Or `PARAMETER` to the Modlefile? Or is there some other similar command? [Here](https://huggingface.co/leafspark/DeepSeek-V2-Chat-GGUF) is a situation that i have to override some parameters. It works at llama.cpp, but i am not sure in ollama. ``` Metadata KV overrides (pass them using --override-kv, can be specified multiple times): deepseek2.attention.q_lora_rank=int:1536 deepseek2.attention.kv_lora_rank=int:512 deepseek2.expert_shared_count=int:2 deepseek2.expert_feed_forward_length=int:1536 deepseek2.expert_weights_scale=float:16 deepseek2.leading_dense_block_count=int:1 deepseek2.rope.scaling.yarn_log_multiplier=float:0.0707 ```
GiteaMirror commented 2026-04-12 13:32:26 -05:00
Author
Owner
Copy Link

@kanishkaditya commented on GitHub (Jan 30, 2025):

hey any way to pass kv overrides to ollama yet?
i have been puzzeled whole day due to this, specifically i want to increase the number of expert but its a kv overide. Not sure how to do it yet.

<!-- gh-comment-id:2624709010 --> @kanishkaditya commented on GitHub (Jan 30, 2025): hey any way to pass kv overrides to ollama yet? i have been puzzeled whole day due to this, specifically i want to increase the number of expert but its a kv overide. Not sure how to do it yet.
GiteaMirror commented 2026-04-12 13:32:27 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Jan 30, 2025):

hey any way to pass kv overrides to ollama yet?

No. My change languished without being merged so I closed it. The runners that are now built for ollama no longer support the --override-kv flag.

<!-- gh-comment-id:2624776156 --> @rick-github commented on GitHub (Jan 30, 2025): > hey any way to pass kv overrides to ollama yet? No. My change languished without being merged so I closed it. The runners that are now built for ollama no longer support the `--override-kv` flag.
GiteaMirror commented 2026-04-12 13:32:27 -05:00
Author
Owner
Copy Link

@ProjectMoon commented on GitHub (Feb 1, 2025):

What is the best way to achieve the overrides then? Is it possible to set values in the config json when importing from safe tensors?

<!-- gh-comment-id:2628816013 --> @ProjectMoon commented on GitHub (Feb 1, 2025): What is the best way to achieve the overrides then? Is it possible to set values in the config json when importing from safe tensors?
GiteaMirror commented 2026-04-12 13:32:28 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Feb 1, 2025):

I expect that modifying the config.json and building a new model would work. The runner architecture that ollama uses is undergoing a lot of changes at the moment, when the transition to the new system is complete there may be an opportunity to add a KV override mechanism. Until then, modifying the model is the only way to achieve this.

<!-- gh-comment-id:2628910810 --> @rick-github commented on GitHub (Feb 1, 2025): I expect that modifying the config.json and building a new model would work. The runner architecture that ollama uses is undergoing a lot of changes at the moment, when the transition to the new system is complete there may be an opportunity to add a KV override mechanism. Until then, modifying the model is the only way to achieve this.
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
amd
api
app
bug
build
cli
cloud
compatibility
context-length
create
docker
documentation
embeddings
feature request
feedback wanted
good first issue
gpt-oss
gpu
harmony
help wanted
image
install
intel
js
launch
linux
macos
memory
mlx
model
needs more info
networking
nvidia
ollama.com
performance
pull-request
Mirrored from GitHub Pull Request
 python
question
registry
rendering
thinking
tools
top
vulkan
windows
wsl
No Label
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#3098
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?